Research map / Point clouds / Point cloud classification
Meta-Sampler: Almost-Universal yet Task-Oriented Sampling for Point Clouds
Ta-Ying Cheng, Qingyong Hu, Qian Xie, Niki Trigoni, Andrew Markham
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train sampler jointly against k=3 frozen task networks instead of one (ablation) | SampleNet · ModelNet40 classification (ratio 8) | accuracy: 83.7% -> 88.0% | Improved |
| train sampler jointly against k=3 task networks instead of one (ablation) | SampleNet · ModelNet40 classification (ratio 64) | accuracy: 54.1% -> 61.6% | Improved |
| joint training instead of single-model training (ablation) | SampleNet · ModelNet40 reconstruction (ratio 8) | Chamfer distance: 3.29 -> 3.05 | Improved |
| joint training instead of single-model training (ablation) | SampleNet · ModelNet40 reconstruction (ratio 64) | Chamfer distance: 4.43 -> 4.31 | Improved |
| joint training instead of single-model training (ablation) | SampleNet · ModelNet40 shape retrieval 20-way (ratio 64) | accuracy: 67.0% -> 71.0% | Improved |
| joint training instead of single-model training (ablation) | SampleNet · ModelNet40 shape retrieval 4-way (ratio 8) | accuracy: 99.6% -> 99.7% | Improved |
| increase the number of task models from 1 to 5 (ablation) | Joint-trained SampleNet · ModelNet40 classification (ratio 32) | accuracy: 80.0% -> 83.4% | Improved |
| increase number of task models k (ablation) | Joint-trained SampleNet · ModelNet40 classification (ratio 32) | accuracy: 81.7% -> 82.8% | Mixed |
| train jointly with several models instead of one (ablation) | SampleNet · ModelNet40 training epochs | epochs to converge: around 60 -> around 40 | Improved |
| start from pretrained meta-sampler instead of from scratch | Joint training · Shape retrieval (ratio 16) | accuracy: 96.7% -> 96.9% | Improved |
| use FPS instead of learned joint sampler is replaced by joint training | FPS · ModelNet40 classification (ratio 8) | accuracy: 70.4% -> 88.0% | Improved |
| add pretrained meta-sampler before joint training | Joint training · Point cloud registration (PCRNet, ratio 32) | rotation error | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SampleNet — SampleNet: Differentiable Point Cloud Sampling (CVPR 2020) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- LTA-PCS: Learnable Task-Agnostic Point Cloud Sampling (CVPR 2024 · similarity 0.90)
- MICAS: Multi-grained In-Context Adaptive Sampling for 3D Point Cloud Processing (CVPR 2025 · similarity 0.88)
- SAMBLE: Shape-Specific Point Cloud Sampling for an Optimal Trade-Off Between Local Detail and Global Uniformity (CVPR 2025 · similarity 0.87)
- GCD-Sampling: A General Cross-scale Decoupled Sampling for Point Cloud (AAAI 2025 · similarity 0.87)
- PointMAC: Meta-Learned Adaptation for Robust Test-Time Point Cloud Completion (NeurIPS 2025 · similarity 0.86)
- PLATYPUS: Progressive Local Surface Estimator for Arbitrary-Scale Point Cloud Upsampling (AAAI 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
